Patients with type 1 diabetes would like to enjoy carefree and active lifestyles, conduct physical activities and exercise programs. An artificial pancreas (AP) that does not necessitate manual inputs such as meal or physical activity information from the patient can accommodate these wishes. Research on the development of AP systems that can control blood glucose levels during the physical activities of patients is important because many patients use daily physical activity as an important element of regulating their blood glucose concentrations and participate in individual and/or group sports that dramatically change their blood glucose levels. Also, the activity patterns in several sports with bursts of high-intensity efforts are similar to children's play and the AP system developed in the proposed work will be more conducive to the carefree play of children. To date, few research groups have attempted to test their AP technologies in an environment of exercise and none have focused on the examination of different types of activity. An AP with properly developed control and hypoglycemia early-warning systems can be safe and effective to use both during and after a variety of types of exercise for patients with type 1 diabetes. This technology will dramatically reduce the number and duration of hypoglycemic events, as compared to the currently available methods of insulin therapy (continuous subcutaneous insulin infusions or multiple daily injections). Such AP systems can only be developed by using a sophisticated multivariable approach that includes glucose concentrations and a number of physiological variables that impact glucose homeostasis such as energy expenditure. Our multivariable recursive modeling and adaptive control framework provides the proper setting to achieve AP systems that are effective during and after several of types of physical activities that differ markedly in energy expenditures and the metabolic systems used to support that expenditure (i.e. aerobic, anaerobic, and mixed activities, team sports). The proposed AP uses patient-specific recursive dynamic models that predict blood glucose concentrations by using subcutaneous glucose measurements and physiological data, early warning systems for hypoglycemia, and adaptive controllers based on these models to calculate insulin infusion rates. The project aims are to develop multivariable control and hypoglycemia early warning systems for APs that will be safe to use during and after various types of exercise and group sports for patients with diabetes, to develop a multivariable simulation system to simulate the effects of different types of exercise on variations in blood glucose levels and test the algorithm developed, to assess the performance of the AP system in clinical studies at Clinical Research Centers and diabetes sports camps and to determine the impact of the AP system developed for changes in fear of hypoglycemia, quality of life, and treatment satisfaction.